A principal component analysis approach to heart rate turbulence assessment in Chagas disease

The analysis of heart rate turbulence (HRT) is a powerful method to estimate the baroreflex from the 24 h Holter ECG signals, by considering that an isolated premature ventricular contraction (PVC) causes an immediate cardiac acceleration followed by a deceleration in normal subjects. This study aims at developing a method for risk stratification of sudden death in chronic Chagas cardiomyopathy, by applying Principal Component Analysis (PCA) to the averaged tachogram segments extracted for HRT analysis. HRT analysis was applied to a database of high resolution ECG from Chagas disease patients, with 10 min signals in three leads, sampled with 16-bit resolution at 1000 Hz. From a set of 115 records that presented premature ventricular contractions (PVC), it was possible to extract at least one valid tachogram for HRT analysis in just 51 signals. The valid segments from each ECG record were taken to compute a coherent mean, used them for measuring the parameters turbulence onset (TO) and turbulence slope (TS). From this dataset, two groups of eight signals were extract, according to the estimated risk of sudden death: high risk (TO ≥ 0 and TS ≤ 2,5 ms/RR interval) and low risk (TO > 0 and TS > 2,5 ms/RR interval). PCA was thus applied to this 16 coherent means of 19 samples to reduce data representation to three principal components (PC), which represented 99.5% of the original variance. Applied to the respective PC scores, a logistic regression allowed the separation of groups with 94% accuracy, 88% sensibility and 100% specificity. As a conclusion, PCA has a potential for baroreflex assessment throughout HRT in Chagas disease, but this method should be validated with a larger sample with long duration ECG.